[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
bug#26497: glibc 2.25 broken on i686
From: |
Ricardo Wurmus |
Subject: |
bug#26497: glibc 2.25 broken on i686 |
Date: |
Mon, 24 Apr 2017 23:24:33 +0200 |
User-agent: |
mu4e 0.9.18; emacs 25.1.1 |
Ludovic Courtès <address@hidden> writes:
> Ricardo Wurmus <address@hidden> skribis:
>
>> Ludovic Courtès <address@hidden> writes:
>>
>>> Ricardo Wurmus <address@hidden> skribis:
>>>
>>>> Guix on i686 is broken since the update to glibc 2.25. There are
>>>> seemingly random segfaults all over the place on my server.
>>>>
>>>> Ludo posted this upstream discussion:
>>>>
>>>> https://sourceware.org/bugzilla/show_bug.cgi?id=21182
>>>
>>> Perhaps we don’t see the problem when running i686 code on x86_64.
>>>
>>>> I could test the patch on my i686 system, but it would take much too
>>>> long for me to build the system from source on my machines.
>>>
>>> Could you first check if there’s a simple way to reproduce it on your
>>> machine? For example, run “guix build coreutils” or “guix build
>>> coreutils --check” and see if that fails.
>>>
>>> If it does, you could arrange to test the patch just in this setup.
>>
>> Do you mean apply the patch and build on i686 or apply the patch and
>> build with “--system” on my x86_64 system?
>>
>> On the i686 system I cannot run anything with the latest glibc, so
>> building there will likely fail unless the bug is fixed.
>
> What I mean is that if you could test the patch in any way you like ;-)
> and ensure that it fixes the problem, it would be great.
Oh, okay. I had misunderstood your message :)
I applied the following patch to my i686 netbook and waited a day for
coreutils and procps to be built with the new glibc. I cannot reproduce
any crashes with these packages, so this seems to be fine.
> Then we soon apply the patch conditionally on glibc (like I did for the
> Coreutils patch for ARM) ASAP so that Hydra has time to rebuild the i686
> world.
The attached patch is for core-updates, but I could also try to make it
apply conditionally for i686.
>From 2b2f1d4947e2198f7011b00a496be078f6a924fd Mon Sep 17 00:00:00 2001
From: Ricardo Wurmus <address@hidden>
Date: Mon, 24 Apr 2017 23:15:41 +0200
Subject: [PATCH] gnu: glibc/linux: Fix segfaults on i686.
* gnu/packages/patches/glibc-memchr-overflow-i686.patch: New file.
* gnu/local.mk (dist_patch_DATA): Add it.
* gnu/packages/base.scm (glibc/linux)[source]: Add patch.
---
gnu/local.mk | 1 +
gnu/packages/base.scm | 1 +
.../patches/glibc-memchr-overflow-i686.patch | 74 ++++++++++++++++++++++
3 files changed, 76 insertions(+)
create mode 100644 gnu/packages/patches/glibc-memchr-overflow-i686.patch
diff --git a/gnu/local.mk b/gnu/local.mk
index a0d7cfd0a..61c3f6ad8 100644
--- a/gnu/local.mk
+++ b/gnu/local.mk
@@ -609,6 +609,7 @@ dist_patch_DATA =
\
%D%/packages/patches/glibc-bootstrap-system.patch \
%D%/packages/patches/glibc-ldd-x86_64.patch \
%D%/packages/patches/glibc-locales.patch \
+ %D%/packages/patches/glibc-memchr-overflow-i686.patch \
%D%/packages/patches/glibc-o-largefile.patch \
%D%/packages/patches/glibc-versioned-locpath.patch \
%D%/packages/patches/gmp-arm-asm-nothumb.patch \
diff --git a/gnu/packages/base.scm b/gnu/packages/base.scm
index 4fd9ed87e..52074c655 100644
--- a/gnu/packages/base.scm
+++ b/gnu/packages/base.scm
@@ -535,6 +535,7 @@ store.")
(modules '((guix build utils)))
(patches (search-patches "glibc-ldd-x86_64.patch"
"glibc-versioned-locpath.patch"
+ "glibc-memchr-overflow-i686.patch"
"glibc-o-largefile.patch"))))
(build-system gnu-build-system)
diff --git a/gnu/packages/patches/glibc-memchr-overflow-i686.patch
b/gnu/packages/patches/glibc-memchr-overflow-i686.patch
new file mode 100644
index 000000000..0b1b5b9f9
--- /dev/null
+++ b/gnu/packages/patches/glibc-memchr-overflow-i686.patch
@@ -0,0 +1,74 @@
+Extracted from glibc upstream git repository. Changes to the ChangeLog have
+been removed. This patch is needed to fix spurious segmentation faults on
+i686.
+
+From 3abeeec5f46ff036bd9df60bb096e20314ccd078 Mon Sep 17 00:00:00 2001
+From: Adhemerval Zanella <address@hidden>
+Date: Tue, 14 Mar 2017 14:16:13 -0300
+Subject: [PATCH] Fix i686 memchr overflow calculation (BZ#21182)
+
+This patch fixes the regression added by 23d2770 for final address
+overflow calculation. The subtraction of the considered size (16)
+at line 120 is at wrong place, for sizes less than 16 subsequent
+overflow check will not take in consideration an invalid size (since
+the subtraction will be negative). Also, the lea instruction also
+does not raise the carry flag (CF) that is used in subsequent jbe
+to check for overflow.
+
+The fix is to follow x86_64 logic from 3daef2c where the overflow
+is first check and a sub instruction is issued. In case of resulting
+negative size, CF will be set by the sub instruction and a NULL
+result will be returned. The patch also add similar tests reported
+in bug report.
+
+Checked on i686-linux-gnu and x86_64-linux-gnu.
+
+ * string/test-memchr.c (do_test): Add BZ#21182 checks for address
+ near end of a page.
+ * sysdeps/i386/i686/multiarch/memchr-sse2.S (__memchr): Fix
+ overflow calculation.
+---
+ string/test-memchr.c | 6 ++++++
+ sysdeps/i386/i686/multiarch/memchr-sse2.S | 2 +-
+ 3 files changed, 15 insertions(+), 1 deletion(-)
+
+diff --git a/string/test-memchr.c b/string/test-memchr.c
+index 2403c9242b..669e092e7d 100644
+--- a/string/test-memchr.c
++++ b/string/test-memchr.c
+@@ -210,6 +210,12 @@ test_main (void)
+ do_test (0, i, i + 1, i + 1, 0);
+ }
+
++ /* BZ#21182 - wrong overflow calculation for i686 implementation
++ with address near end of the page. */
++ for (i = 2; i < 16; ++i)
++ /* page_size is in fact getpagesize() * 2. */
++ do_test (page_size / 2 - i, i, i, 1, 0x9B);
++
+ do_random_tests ();
+ return ret;
+ }
+diff --git a/sysdeps/i386/i686/multiarch/memchr-sse2.S
b/sysdeps/i386/i686/multiarch/memchr-sse2.S
+index 910679cfc0..e41f324a77 100644
+--- a/sysdeps/i386/i686/multiarch/memchr-sse2.S
++++ b/sysdeps/i386/i686/multiarch/memchr-sse2.S
+@@ -117,7 +117,6 @@ L(crosscache):
+
+ # ifndef USE_AS_RAWMEMCHR
+ jnz L(match_case2_prolog1)
+- lea -16(%edx), %edx
+ /* Calculate the last acceptable address and check for possible
+ addition overflow by using satured math:
+ edx = ecx + edx
+@@ -125,6 +124,7 @@ L(crosscache):
+ add %ecx, %edx
+ sbb %eax, %eax
+ or %eax, %edx
++ sub $16, %edx
+ jbe L(return_null)
+ lea 16(%edi), %edi
+ # else
+--
+2.12.2
+
--
2.12.2
--
Ricardo
GPG: BCA6 89B6 3655 3801 C3C6 2150 197A 5888 235F ACAC
https://elephly.net